def test_normalize_nonlinpart_go():
"""Manual tests for normalize_nonlinpart_go"""
gasoil = GasOil(swl=0.1, sgcr=0.12, sorg=0.05, h=0.05)
gasoil.add_corey_gas(ng=2.1, krgend=0.9)
gasoil.add_corey_oil(nog=3, kroend=0.8)
krgn, kron = normalize_nonlinpart_go(gasoil)
assert np.isclose(krgn(0), 0)
assert np.isclose(krgn(1), 0.9)
# kron is normalized on son
assert np.isclose(kron(0), 0)
assert np.isclose(kron(1), 0.8)
# Test with tricky endpoints
h = 0.01
gasoil = GasOil(swl=h, sgcr=h, sorg=h, h=h)
gasoil.add_corey_gas(ng=2.1, krgend=0.9)
gasoil.add_corey_oil(nog=3, kroend=0.8)
krgn, kron = normalize_nonlinpart_go(gasoil)
assert np.isclose(krgn(0), 0.0)
assert np.isclose(krgn(1), 0.9)
assert np.isclose(kron(0), 0)
assert np.isclose(kron(1), 0.8)
# Test again with zero endpoints:
gasoil = GasOil(swl=0, sgcr=0, sorg=0, h=0.01)
gasoil.add_corey_gas(ng=2.1, krgend=0.9)
gasoil.add_corey_oil(nog=3, kroend=0.8)
krgn, kron = normalize_nonlinpart_go(gasoil)
assert np.isclose(krgn(0), 0.0)
assert np.isclose(krgn(1), 0.9)
assert np.isclose(kron(0), 0)
assert np.isclose(kron(1), 0.8)
# Test when endpoints are messed up (cleared)
gasoil = GasOil(swl=0.1, sgcr=0.2, sorg=0.1, h=0.1)
gasoil.add_corey_gas(ng=2.1, krgend=0.6)
gasoil.add_corey_oil(nog=3, kroend=0.8)
gasoil.swl = 0
gasoil.sgcr = 0
gasoil.sorg = 0
krgn, kron = normalize_nonlinpart_go(gasoil)
# These go well still, since we are at zero
assert np.isclose(krgn(0), 0.0)
assert np.isclose(kron(0), 0)
assert np.isclose(kron(1), 0.8)
# These do not match when endpoints are wrong
assert not np.isclose(krgn(1), 0.6)
# So fix endpoints!
gasoil.swl = 1 - gasoil.table["sg"].max()
gasoil.sgcr = gasoil.estimate_sgcr()
gasoil.sorg = gasoil.estimate_sorg()
# Try again
krgn, kron = normalize_nonlinpart_go(gasoil)
assert np.isclose(krgn(0), 0.0)
assert np.isclose(kron(0), 0)
assert np.isclose(krgn(1), 0.6)
assert np.isclose(kron(1), 0.8)
def test_interpolations_go_fromtable():
"""Test based on bug exposed in pyscal 0.6.1, where sgcr
was underestimated in interpolations following add_fromtable().
"""
base = pd.DataFrame(
columns=["Sg", "krg", "krog"],
data=[
[0.0, 0.0, 1.0],
[0.1, 0.0, 1.0],
[0.2, 0.0, 1.0], # sgcr
[0.3, 0.1, 0.9],
[0.8, 0.8, 0.0], # sorg
[0.9, 0.9, 0.0],
[1.0, 1.0, 0.0],
],
)
opt = pd.DataFrame(
columns=["Sg", "krg", "krog"],
data=[
[0.0, 0.0, 1.0],
[0.1, 0.0, 1.0],
[0.3, 0.0, 1.0],
[0.4, 0.1, 0.2], # sgcr
[0.9, 0.9, 0.0], # sorg
[0.95, 0.95, 0.0],
[1.0, 1.0, 0.0],
],
)
go_base = GasOil(h=0.01)
go_base.add_fromtable(base)
assert np.isclose(go_base.estimate_sgcr(), 0.2)
assert np.isclose(go_base.estimate_sorg(), 0.2)
go_opt = GasOil(h=0.01)
go_opt.add_fromtable(opt)
assert np.isclose(go_opt.estimate_sgcr(), 0.3)
assert np.isclose(go_opt.estimate_sorg(), 0.1)
go_ip = interpolate_go(go_base, go_opt, 0.5, h=0.01)
assert np.isclose(go_ip.estimate_sgcr(), 0.25)
assert np.isclose(go_ip.estimate_sorg(), 0.15)
def test_sorg():
"""Test estimate_sorg for some manually set up cases"""
# sorg, sgcr, h, swl:
testtuples = [
(0.3, 0.01, 0.1, 0.1),
(0.2, 0, 0.05, 0.0),
(0.1, 0.3, 0.01, 0.5),
(0.0, 0, 0.1, 0.000001),
(0.9, 0.000001, 0.1, 0),
(0.4, 0, 0.1, 0.2),
]
for testtuple in testtuples:
real_sorg = testtuple[0]
real_sgcr = testtuple[1]
h = testtuple[2]
swl = testtuple[3]
go = GasOil(sgcr=0.03, sorg=real_sorg, h=h, swl=swl)
go.add_corey_oil(nog=2)
go.add_corey_gas(ng=2, krgend=0.9)
print("Testing sorg={} on h={}, swl={}".format(real_sorg, h, swl))
est_sorg = go.estimate_sorg()
mis = abs(est_sorg - real_sorg)
if mis > 0.01:
print("Missed, estimated was {}".format(est_sorg))
assert mis < h + epsilon # Can't guarantee better than h.
# If krgendanchor is not sorg (default), then krg cannot be used
# and the GasOil object will resort to using krog. Should work
# when now=2 but might not always work for all kinds of LET parameters.
go = GasOil(sorg=real_sorg, h=h, swl=swl, krgendanchor="")
go.add_corey_oil(nog=2)
go.add_corey_gas(ng=2, krgend=0.8)
est_sorg = go.estimate_sorg()
print("Estimated to {}".format(est_sorg))
mis = abs(est_sorg - real_sorg)
assert mis < h + epsilon
# Test sgcr:
go = GasOil(sorg=real_sorg,
sgcr=real_sgcr,
h=h,
swl=swl,
krgendanchor="sorg")
go.add_corey_oil(nog=2, kroend=0.8)
go.add_corey_gas(ng=2, krgend=0.8)
est_sgcr = go.estimate_sgcr()
mis = abs(est_sgcr - real_sgcr)
assert mis < h + epsilon
def test_interpolate_go(
swl,
sgcr,
dsgcr,
dswlhigh,
sorg,
dsorg,
ng_l,
ng_h,
nog_l,
nog_h,
krgend_l,
krgend_h,
kroend_l,
kroend_h,
):
# pylint: disable=too-many-arguments,too-many-locals
"""Test many possible combinations of interpolation between two
Corey gasoil curves, looking for numerical corner cases"""
h = 0.01
go_low = GasOil(swl=swl, sgcr=sgcr, sorg=sorg, h=h)
go_high = GasOil(
swl=swl + dswlhigh, sgcr=sgcr + dsgcr, sorg=max(sorg - dsorg, 0), h=h
)
go_low.add_corey_gas(ng=ng_l, krgend=krgend_l)
go_high.add_corey_gas(ng=ng_h, krgend=krgend_h)
go_low.add_corey_oil(nog=nog_l, kroend=kroend_l)
go_high.add_corey_oil(nog=nog_h, kroend=kroend_h)
ips = []
ip_dist = 0.05
for t in np.arange(0, 1 + ip_dist, ip_dist):
go_ip = interpolate_go(go_low, go_high, t)
check_table(go_ip.table)
ips.append(go_ip)
assert 0 < go_ip.crosspoint() < 1
# sgcr is non-trivial, if exponents are high, an effective sgcr might
# be larger than the value used to initialize the curve. Try to be
# permissive enough here. This can even cause the interpolant to be
# outside the low-high envelope, but it is the way it is supposed to
# be when sgcr is interpolated separately.
sgcr_low = min(
go_low.sgcr, go_low.estimate_sgcr(), go_high.sgcr, go_high.estimate_sgcr()
)
sgcr_high = max(
go_low.sgcr, go_low.estimate_sgcr(), go_high.sgcr, go_high.estimate_sgcr()
)
sgcr_ip = go_ip.estimate_sgcr()
sgcr_lower_bound_ok = sgcr_low - h - epsilon < sgcr_ip
sgcr_upper_bound_ok = sgcr_ip < sgcr_high + h + epsilon
assert sgcr_lower_bound_ok
assert sgcr_upper_bound_ok
# Distances between low and interpolants:
dists = [
(go_low.table - interp.table)[["krg", "krog"]].sum().sum() for interp in ips
]
print(
"Interpolation, mean: {}, min: {}, max: {}, std: {} ip-par-dist: {}".format(
np.mean(dists), min(dists), max(dists), np.std(np.diff(dists[1:])), ip_dist
)
)
assert np.isclose(dists[0], 0) # Reproducing go_low
# All curves that are close in parameter t, should be close in sum().sum().
# That means that diff of the distances should be similar,
# that is the std.dev of the distances is low:
ip_dist_std = np.std(
np.diff(dists[1:])
) # This number depends on 'h' and 't' range, and
# by how different the low and high is.
# (avoiding the first which reproduces go_low
if ip_dist_std > 1.0: # number found from trial and error.
print("ip_dist_std: {}".format(ip_dist_std))
print(dists)
_, mpl_ax = plt.subplots()
go_low.plotkrgkrog(mpl_ax=mpl_ax, color="red")
go_high.plotkrgkrog(mpl_ax=mpl_ax, color="blue")
for interp in ips:
interp.plotkrgkrog(mpl_ax=mpl_ax, color="green")
plt.show()
assert False
